The incomplete CST walker coverage issue has been fixed by adding
general multi-alt group support. All three patterns (repeated
separators, token-only groups, multi-element groups) now produce
stored fields that the walker can traverse.

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC

Replace content-based naming (e.g., KInsertOrKReplaceOrKInsertK...Group)
with concise rule-name+index naming (e.g., InsertStmtGroup0, group_0).

Changes across 5 files:
- gen_util.wado: general_group_type_name(rule_name, group_index),
  general_group_field_name(group_index), general_group_alt_struct_name(rule_name, group_index, alt_index)
- gen_context.wado: add current_rule_name and group_counter fields to GenContext
- generator.wado: thread rule_name and group_counter through type generation
- parser_gen.wado: use ctx.current_rule_name and ctx.group_counter with save/restore
- visitor_gen.wado: thread rule_name and group_counter parameters

Note: triggers a cross-module codegen bug in the Wado compiler when linking
via main.wado (type mismatch: expected (ref $type), found (ref $type)).
Each file compiles individually without errors. Needs compiler fix first.

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC

Three interconnected issues fixed:

1. Group counter divergence between type gen, parser gen, and visitor gen.
   Non-storable groups (groups that don't match any specialized pattern) contain
   nested general groups that advance the counter in the type gen's
   gen_element_variant_types path, but not in collect_fields/element_to_field.
   Fixed by advancing the counter using count_element_groups for non-storable
   groups in all three passes.

2. Shared single-alt group type naming. When two rules (e.g., select_or_values
   and select_core) share the same single-alt group struct (e.g., KFromItemItem),
   the nested general group types were emitted only under the first rule's name.
   The parser gen for the second rule would reference non-existent types.
   Fixed by tracking which rule "owns" each single-alt group's types via
   GenContext.single_alt_owner map, and overriding current_rule_name in
   gen_consume_single_alt_group.

3. Compiler arity check for cross-module function calls was missing, causing
   confusing codegen errors instead of proper diagnostics.

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC

Add a check in resolve_struct_literal that all struct fields must be
provided in the literal. Previously, omitting a field silently produced
invalid Wasm (struct.new with wrong number of values on the stack),
causing runtime type mismatches like "expected (ref $type), found
(ref (exact $type))". Now the compiler emits a clear "missing field"
compile error.

Also adds e2e tests for missing fields, extra fields, and
cross-module struct field types.

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC

Distinguish between "field not found" (field access like x.foo) and
"extra field in struct literal" (struct literal with a field that
doesn't exist on the struct). The new error message is:
  struct 'Foo' has no field 'bar'

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC

gen_general_alt_parse_and_store and gen_general_alt_construct used
gen_element for each element, which calls gen_repeat (greedy) for
Repeat nodes. This lost the non-greedy while-loop guard (peek_at(1)
disambiguation) that gen_elements_with_non_greedy provides.

The regression caused CREATE TABLE with multiple columns and table
constraints to fail parsing because the comma_column_def loop consumed
commas greedily instead of stopping at table constraints.

Add failing test first (TDD), then fix by using
gen_elements_with_non_greedy in general group code generation.

https://claude.ai/code/session_01VvNqsUaUQNBJkaUzWQFuZC